File size: 7,101 Bytes
0d25dbe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
# PhoneWatchUI

This dataset is for research purposes only.

## Dataset structure

```text
dataset/
├── apple_store/
│   ├── dataset/
│   │   ├── apps.jsonl
│   │   ├── images.jsonl
│   │   └── images/<app_id>/{watch,phone}/<position>.png
│   └── script/
│       ├── app_collect.py
│       ├── watch_app_filter.py
│       ├── image_download.py
│       └── README.md
├── google_play/
│   ├── dataset/
│   │   ├── apps.jsonl
│   │   ├── images.jsonl
│   │   └── images/<app_folder>/<position>.<extension>
│   └── script/
│       ├── app_collect.py
│       ├── watch_app_filter.py
│       ├── image_download.py
│       ├── device_filter.py
│       ├── requirements.txt
│       └── README.md
├── README.md
└── requirements.txt
```

Each store has its own app records, image records, and image files. `image_path` is relative to that store's `dataset/` directory. Google Play folder names preserve distinctions between case-sensitive app IDs.

## Fields

JSONL files are UTF-8, with one JSON object per line. Each app has one app record; each gallery position has one image record. Repeated URLs and image bytes are retained. Missing source attributes may be absent or null.

### `apps.jsonl`

| Field | Description |
| --- | --- |
| `app_id` | Apple numeric App ID or Google Play package ID, stored as a string. |
| `store` | `apple_store` or `google_play`. |
| `title` | App name. |
| `description` | App description. |
| `source_url` | App listing URL. |
| `metadata` | Store-specific app attributes below. |

**Apple App Store:** `metadata.watch_manifest` and `metadata.phone_manifest` each contain `genres`, `screenshots`, and `metadata`. Screenshot entries contain `rank`, `url`, `width`, `height`, and `platform`. The nested `metadata` fields are:

| Fields | Description |
| --- | --- |
| `developer_name`, `developer_id`, `developer_url`, `website_url` | Developer identity and links. |
| `price`, `currency`, `is_free` | Pricing. |
| `rating_average`, `rating_count` | Ratings. |
| `app_version`, `app_version_updated_at` | Version and update date. |
| `compatibility`, `platforms` | Supported platforms; each compatibility entry has `platform` and `requirement`. |

**Google Play:** `metadata` contains the following attributes. BSON-specific values retain MongoDB Extended JSON objects.

| Fields | Description |
| --- | --- |
| `summary` | Short app description. |
| `developer`, `developerId`, `developerInternalID`, `developerWebsite` | Developer identity and website. |
| `genre`, `genreId`, `categories` | App categories. |
| `price`, `currency`, `free` | Pricing. |
| `score`, `ratings` | Rating score and count. |
| `version`, `updated`, `androidVersion`, `androidVersionText`, `androidMaxVersion` | App version, update date, and Android requirements. |
| `screenshots`, `icon`, `headerImage`, `videoImage` | Store media URLs or entries. |
| `permissions`, `datasafety` | Permissions and data-safety disclosures. |

### `images.jsonl`

| Field | Description |
| --- | --- |
| `image_id` | Position identity: `<store>:<app_id>:<gallery>:<position>`. |
| `app_id`, `store` | Join key to the corresponding app record. |
| `gallery` | Apple: `watch` or `phone`; Google Play: `screenshots`. |
| `position` | Zero-based gallery position. |
| `image_path` | Relative file path; null when no file is available. |
| `width`, `height` | Image dimensions in pixels; null if unavailable. |
| `aspect_ratio` | Width divided by height; null if unavailable. |
| `device` | `watch`, `phone`, or `other`. |
| `device_label_source` | Labeling method and parameters, described below. |
| `source_url` | Source image URL. |
| `sha256` | Recorded SHA-256 of the image bytes; null if unavailable. |

Apple labels follow the source gallery and use `{"method":"explicit_app_store_gallery"}`. Google Play uses `{"method":"aspect_ratio","phone_ratio":1.5}`: square images are `watch`, non-square images with long/short side ratio ≥ 1.5 are `phone`, and remaining or dimensionless images are `other`. These labels identify the source gallery or image shape, not necessarily the devices depicted.

## Image download

Download images using the `source_url` in each `images.jsonl` record. For Google Play, request the original-resolution URL with `=s0`, as implemented in `image_download.py`. Compute SHA-256 over the downloaded file bytes before resizing or re-encoding, and compare it with the record's `sha256`; a match confirms byte-for-byte identity with the released image. Records with a null hash cannot be verified this way, and changed or unavailable source URLs may prevent retrieval of matching bytes.

## Script usage

Use Python 3.10+. Run the commands below from the dataset root, with new outputs under `work/`. Keep each store's scripts together. Apple scripts use the standard library; install Google Play dependencies with:

```sh
python3 -m pip install -r requirements.txt
```

### Apple App Store

Prepare `apple_ids.txt` with one numeric App ID or App Store URL per line. The three steps collect app metadata, select apps with explicit Watch platform support, and download their Watch/iPhone galleries:

```sh
python3 apple_store/script/app_collect.py --app-ids apple_ids.txt --output work/apple/candidates.jsonl
python3 apple_store/script/watch_app_filter.py --input work/apple/candidates.jsonl \
  --output work/apple/dataset/apps.jsonl
python3 apple_store/script/image_download.py --apps work/apple/dataset/apps.jsonl --output work/apple/dataset
```

### Google Play

Prepare `google_ids.txt` with one package ID per line. The four steps collect app metadata, select watch-related apps, download gallery images, and assign device labels:

```sh
python3 google_play/script/app_collect.py --ids google_ids.txt --output work/google/candidates.jsonl
python3 google_play/script/watch_app_filter.py --input work/google/candidates.jsonl \
  --output work/google/dataset/apps.jsonl
python3 google_play/script/image_download.py --apps work/google/dataset/apps.jsonl --output work/google/dataset
python3 google_play/script/device_filter.py --input work/google/dataset/images.jsonl \
  --output work/google/dataset/images.jsonl --phone-ratio 1.5
```

`watch_app_filter.py` matches Wear OS, Android Wear, smartwatch, watch face, and watch app/application/version in `title`, `summary`, or `description`. Exact patterns are defined in the script.

Google Play `image_download.py` produces image records without `device` and `device_label_source`; `device_filter.py` adds them. Its default retains every record and supports replacing the input JSONL atomically. To export a subset, use `--device watch`, `--device phone`, or `--device other` with a separate output path; image files are unchanged.

Both stores write `apps.jsonl`, `images.jsonl`, and `images/` in the download output directory. Both app collectors also accept app JSONL as the ID input. Use `--help` on any script for its remaining options.